Industries / Energy, Oil & Gas · Distributed energy · Upstream & oilfield services
Two halves of one problem. In energy, the device knows its own state, the telemetry lands somewhere else, and the customer report is assembled by hand from a weekly CSV. In oil & gas, production lives in a historian, costs live in WolfePak, Quorum or P2, and lease terms are locked inside scanned PDFs. Either way the translation between systems is a person with a spreadsheet. We build the layer that removes that step — and the AI layer that reads the documents nobody has time to read.
You keep the one-page data flow diagram either way. No obligation attached to it.
Both halves, one problem shape
Nothing here is exotic. Each of these was bought for a good reason, by someone who was right at the time — and none of them was bought to talk to the others.
The gap, measured
Scrub through a month and watch a single artificial lift failure travel between your systems. Then switch the panel to a connected state and run the same month again.
Where the data actually starts
Every one of these builds begins at the edge — a device, a sensor, a controller — and the interesting engineering is what happens between there and the report someone reads on Monday.
Where the day goes
Field data capture
A pumper finishes a route at 14:30 and leaves eight paper tickets on a desk, or texts photographs of them. Someone keys the gauge readings and run tickets into Excel the following morning. By the time a number reaches anyone able to act on it, the well has been doing whatever it has been doing for two days.
Downtime attribution
The historian recorded the artificial lift failure the moment the pump card went flat. The LOE line that reflects it appears at month-end close, three weeks later. By then downtime attribution is a conversation between two people with two different spreadsheets, and the data cannot settle it.
Document intelligence
There is an answer. It is inside several thousand scanned PDFs and TIFFs of leases, JOAs and division orders, and reaching it means a landman with a highlighter working most of a week. The same question next quarter costs the same week again.
Regulatory
Form PR, W-10 and G-1 for the Texas RRC, state-by-state production reporting, FracFocus disclosures, and Subpart W volumes for the EPA GHGRP — all assembled by hand from three systems every month, then checked twice because last quarter a decimal moved one place.
JIB & division orders
A JIB reconciliation dispute, or a decimal interest on a division order that no longer matches the recorded assignment. The discrepancy sat in the data for months. Nothing was watching for it, because watching for it means holding land records and accounting in the same place.
Analytics
Cost per BOE by lease. Decline curve analysis against actuals, with LOE alongside. Any question that touches production data and financial data at once begins with exporting two files and rebuilding the same spreadsheet somebody already built last quarter.
Software close to hardware
We rebuilt the C++ control software and operator interface for machines that bore through rock. Software that sits this close to expensive equipment is the same category of work as a microgrid controller or a plunger-lift controller.
The energy half
Solar and storage hardware makers, C&I owner-operators, microgrid and EV-charging operators, utility innovation groups, and energy-hardware companies with a working device and no software team. The engineering org is weighted toward hardware, and the roadmap keeps arriving as firmware, cloud and an app at the same time.
Device
Zephyr, FreeRTOS, ESP-IDF or bare metal on STM32-class parts, with secure boot, signed images, A/B partitions and a rollback that works when an OTA campaign meets a bad connection. Battery management, state of charge and health, cell balancing, thermal derating and MPPT charge control where the device does its own thinking.
Telemetry
Modbus RTU/TCP and CAN at the device, MQTT with Sparkplug B on the way up, an edge gateway that buffers and forwards when the link returns. Then the part everyone underestimates: time-series storage with downsampling, gap-filling, late-arriving data, and unit and timezone normalisation that survives a fleet in three countries.
Grid
IEC 61850, DNP3, IEEE 2030.5 and OpenADR for demand response and smart-inverter control; DER aggregation and virtual-power-plant dispatch with the telemetry a CAISO or ERCOT programme expects. Interconnection behaviour under rules like California Rule 21 and UL 1741 SB / IEEE 1547-2018 is a design constraint we work to, not a certification we hold.
Charging
OCPP 1.6J and 2.0.1 against a CSMS, OCPI for roaming, and session-to-CDR reconciliation — because the argument is never about the kilowatt-hours, it is about which record the money follows.
Reporting
Energy performance reporting, ISO 50001-style measurement and verification, emissions accounting, and the Scope 2 distinction between market-based and location-based that decides whether your report is comparable to anyone else's. RECs and hourly matching where the customer has committed to it.
The manual step
Somebody exports the fleet, opens a spreadsheet, and rebuilds the customer report. It is the same person, the same afternoon, every week — and it is the step we remove first, because it is the one that stops being possible the moment the fleet doubles.
What we build
Timelines below are what we would commit to for a first phase after discovery, on a typical single-operator environment. They get firmer, not vaguer, once we have seen your systems.
01
One queryable source of truth across historian, accounting and land, so a question crosses systems without a person in the middle.
02
Lease terms, royalty and working-interest decimals, Pugh clauses, continuous-operations provisions and expiry dates, pulled out of scanned documents into a database you can query.
03
Pumper tickets captured on a phone at the lease, in a section with no signal, and reconciled the moment the truck reaches coverage.
04
Ask production and financial data a question in plain English and get the number, instead of commissioning a report.
05
Artificial lift failure, tubing leaks and unplanned downtime flagged in minutes from the tags you already historize.
06
Recurring filings assembled and validated from source data, on a schedule, with the variances surfaced before submission.
Hardware companies buy software from people who have stood next to the machine.
How an engagement runs
Step 00
A call with the engineers, not a salesperson. We map every place data moves between your systems and mark the handoffs done by hand. You keep the one-page diagram.
Free · 30 minutesStep 01
Paid, time-boxed, defined deliverable: system access review, integration inventory, and a written build plan with a fixed price on phase one. You own everything it produces whether or not we continue.
Paid · ~2 weeksStep 02
One narrow piece in production and in use — a single integration, a single filing, one document type. You judge working software rather than a status report.
4–6 weeksStep 03
Each subsequent phase is scoped against what the last one proved, so scope grows from measured results rather than from an original guess.
Phase by phaseStraight answers
Correct, and we have no interest in replacing Quorum, WolfePak or your historian. Those systems are good at what they do and your team knows them. What is missing is the layer between them, and we work alongside your existing vendors to build it. The domain knowledge that matters for that job is how your data actually moves — which is exactly what the audit establishes before anyone writes a line of code.
Deployment goes into your VPC or on-premise, so operational data stays inside your boundary. Your data is never used to train models, and any LLM processing runs against providers under zero-retention terms. The engineers on your project are named individuals under NDA, working on least-privilege read access that you grant and can revoke. We do not hold SOC 2 or ISO 27001 certification — if your process requires a certified vendor, tell us at the audit and we will say plainly whether we are a fit.
The common thread in those stories is an open-ended build against a target that kept moving, billed monthly, with the first usable output arriving nine months in. That is why the entry point here is a paid, fixed-scope discovery with a written plan and a fixed price attached to phase one, and why the first working slice lands in weeks. You own the output at every stage — schema, source code, documentation, credentials — so even the version of this where you walk away leaves you holding a map of your own data.
Our working day runs 10:00–19:00 CET, which puts a live window against every US morning, and a written handoff is waiting when your day starts. Overtime is capped and cannot be billed without your written approval in advance, so nobody quietly runs up hours against a deadline. Fluvius USA Inc is a US company, incorporated in California, and you contract with it under US law. We have been shipping production software since 2017, including work delivered on-site inside a retail bank and applications live in production for government ministries. The team works European hours with full US-morning overlap, so you have live hours with the engineers on your project every working day, and you talk to them directly rather than through an account manager.
The first working deliverable lands in weeks. The phasing exists precisely so you are never asked to fund a year on faith: audit, then a two-week paid discovery, then one narrow slice in production inside 4–6 weeks. If the first slice does not earn its keep, you have spent a small, defined amount and you own what was built.
You do, at every stage: source, documentation, infrastructure and credentials. The transfer is worldwide, with no time limit and no territorial limit, and anyone we bring onto the work assigns their intellectual property to you in writing before they touch it — the clause most agreements leave out, and the one that matters when the asset changes hands.
The first 30 days of any engagement are a trial period, and either side can end it in that window without notice and without a penalty. It is in the agreement rather than in a sales conversation. A month of real work on your data tells both of us more than another round of evaluation would.
Where our experience transfers
We have not yet shipped a named upstream oil & gas product, and we would rather say so than let a skimming reader conclude otherwise. That half of the page argues from domain fluency and transferable capability. The energy half does not have to: the Sonnenrepublik programme is ours, it is named, and it has run for years.
Industrial control · C++
Modernised the C++ control software and operator interface for tunnel-drilling machines: manual operation down 90%, efficiency up 240%. Machine control, operator HMI, and software that has to behave predictably around expensive equipment.
Sensor fusion · real time
Fused acoustic, RF, EO/IR and radar inputs into a single validated track, moving from detection to handoff in four seconds. Multi-source time-series reconciliation, which is the same problem shape as reconciling historian tags against field reports.
Direct energy case · named
Firmware, embedded software and hardware R&D for a novel power station for PV systems, built and sustained across years of product evolution. Not a dashboard bolted onto somebody else's hardware: the software that runs the device, the telemetry that leaves it, and the application people use — from one team, as one system. This is the case that makes the energy half of this page real.
Operations automation · LLM
A pipeline moving records across CRM, email, project management and reporting with LLM classification and enrichment: 300+ records a week automated, and roughly 80% less manual data entry in the first month.
What the firmware runs on
The energy case above is not a dashboard over somebody else’s hardware. This is the enclosure the firmware ships inside — designed, built and brought up as one system with the software that runs it.
Who runs the audit
The Data Flow Audit is run by an engineer who has worked on this kind of system, not by a sales team — thirty minutes, one diagram, and no obligation attached to it.
The record behind the claim
Hours and delivered value come straight out of the system we run every project through. The Upwork figures next to them you can verify yourself in about thirty seconds.
Clients served
Years of happy clients
Years our longest clients have stayed
Job Success on Upwork
★★★★★ 5.0
“They understood a difficult brief and delivered everything professionally, on time and on budget. I hope we have the opportunity to work together again soon.”
★★★★★ 5.0
“Clean code, clear documentation, great communication and able to work well with a minimal brief and a constantly changing UX design.”
Many of our clients have been with us for 7+ years straight. 100% Job Success on Upwork · 4.9 / 5 on Clutch. Verify →
Not ready to talk yet
A one-page diagnostic you can run against your own operation in about ten minutes. If you recognise more than four of them, the audit is worth half an hour of your time. If you recognise none, we would genuinely like to know how you got there.
One email with the checklist attached. No sequence, no newsletter.
Primary next step
Thirty minutes with the engineers who would do the work. We map how data moves between your systems today and mark every handoff that runs on a person. You keep the diagram regardless of what happens next.
What you walk away with
The diagram is yours. Use it with us, with your existing vendor, or with nobody.
